`TypeSize` is not granular enough

[onbjerg]

Reproducible steps

TypeSize stores ceil(bit_len/8), so int_literal[1] covers the range 0-255.

This means we can't distinguish edge cases like 127 (fits in int8) from 128 (doesn't fit), which is problematic for implicit and explicit integer conversions.

See #633 (comment) for additional discussion.

Solar version

N/A

[dipanshuhappy]

Hi @onbjerg , happy to work on this. I took the time to understand the issues. I am thinking of approaching it the following ways

1. Storing directly bits rather than byte.
   At

/// Byte size of a fixed-bytes, integer, or fixed-point number (M) type. Valid values: 0..=32.
#[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
pub struct TypeSize(u8);

Since we kinda ceil using the original bit value. Why not just store the bit value.
In terms of refactoring, its gonna be a pain but granularity would be achieved. Hence we can imagine

/// Bit size of a fixed-bytes, integer, or fixed-point number (M) type. Valid values: 0..=256.
#[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
pub struct TypeSize(u16);

2. Patching just for int literal conversions
   Since

IntLiteral(bool,TypeSize)

We can change this to its own struct which tracks the bit precisely.
Major con here would be that it would bite us somewhere else, as the core bug is still not solved.

[DaniPopes]

let's go with the first one

[onbjerg]

Apologies @dipanshuhappy, I'm taking this over since this is blocking future work atm